Sketches: Normand Daigle
After outlining some technical concepts in the first two articles of this series on power amplifier families, last month we explored amplifiers operating in Class A. Despite their undeniable qualities, they have a significant and unavoidable drawback: massive energy consumption at all times, even when no audio signal is present, along with the inherent thermal dissipation caused by their operating mode. While performance is excellent, with an efficiency rate of about 30% and very high associated manufacturing costs, it’s clear that a more efficient configuration is desirable. The solution was to create a push-pull output stage where the quiescent current is zero in the case of Class B, or very low for Class AB, drastically reducing the device’s power consumption.
Class B is rarely used today due to its weaknesses, which we will discuss, but understanding its basic concept helps in comprehending Class AB. The vast majority of power amplifiers currently on the market operate in Class AB, including those with enviable reputations. This system has merit but also presents challenges that must be addressed to achieve the desired result. The basic concept is relatively simple: the upper half of the output circuit handles the positive half of the wave, while the lower half takes care of the negative part, as illustrated in Sketch 1. The availability of complementary polarity semiconductors (NPN and PNP for conventional transistors, or N-channel and P-channel for FET and MOSFET transistors) allows for the creation of a simple and elegant output circuit. It is possible to design a quasi-complementary Class B or AB output stage with same-polarity transistors, but such a configuration offers no advantages and even adds to circuit complexity. Therefore, we will only briefly mention it here, as it is no longer used today.
Push-Pull Output Stage in Class B
The Class B circuit shown above can work for applications with low-quality requirements but would fall short for HiFi applications. In practice, the waveform appearing at the speaker output connection would be affected by significant crossover distortion. This phenomenon occurs because the base-emitter junction of a bipolar junction transistor (BJT) must reach about 0.6 V before current can flow between the collector and emitter. The first 0.6 V of each half of the waveform is therefore truncated, resulting in what is seen in Figure 2, an obviously unacceptable situation for high-quality reproduction. Nevertheless, the circuit offers the advantage of very low power consumption at rest.
Illustration of Crossover Distortion
A way to minimize crossover distortion without reducing the circuit’s energy efficiency is to apply feedback, which involves taking a sample of the output signal and returning it to the amplifier’s inverting input for correction by comparing the input signal with the output. Unfortunately, this approach tends to create more problems than it solves. To achieve an acceptable level of distortion, the feedback rate must be very high, requiring a high open-loop gain (without feedback), which significantly impacts overall performance. The resulting circuit would need heavy compensation to ensure unconditional stability with the complex load (resistive-inductive-capacitive) presented by a speaker. While static measurements of such an amplifier may appear promising, dynamic performance during musical reproduction is much less impressive. The following diagram illustrates such a configuration, where the overall amplifier gain equals (R2 / R1) + 1, typically around 20 to 30 dB.
Class B Amplifier with Feedback
A more effective way to control distortion without overly compromising efficiency or relying on very high feedback levels is to use a method to bridge the 0.6 V base-emitter junction offset. In this case, each output transistor permanently receives the voltage required to reach or slightly exceed its threshold. In the circuit below, a 1.2 V source is connected to the bases of the output transistors. This slight biasing creates what is called Class AB, essentially a Class B circuit pushed slightly toward Class A. This approach can completely eliminate crossover distortion while slightly increasing power consumption. The static current of the output stage created by this biasing, also called idle current, is typically about 15 mA per pair for bipolar output transistors (some amplifiers use multiple pairs in parallel to increase power capacity). For a typical 100 W RMS amplifier at 8 Ω with a symmetrical ±40 V power supply, the idle power consumption will be only 1.2 W (15 mA × 80 V, the sum of the positive and negative supplies). For the same available output power, a Class A amplifier would consume about 350 W continuously, clearly illustrating the energy advantage of Class AB. It is worth noting that amplifiers with MOSFETs at the output will have much higher idle currents, around 100 to 150 mA, resulting in greater thermal dissipation.
Amplifier in Class AB
A Class AB amplifier is inherently more linear (i.e., free from distortion) than an equivalent Class A amplifier and thus requires a lower open-loop gain, which improves circuit stability. Unfortunately, the reality tends to be more complex than one might hope. The biasing of the output stage is typically created by a variable circuit that allows for the adjustment of the idle current to achieve the optimal result. This adjustment compensates for variations in the characteristics of the output transistors due to production tolerances. Once the bias is set to the desired idle current, a fixed and stable voltage is applied between the transistor bases.
Since we are dealing with a power circuit, the output transistors could eventually reach relatively high temperatures during high-volume listening, and this is where complications arise. The core issue is that their base-emitter voltage decreases with rising temperature, meaning it has a negative temperature coefficient. Since the voltage between the bases is kept stable by the bias circuit, the idle current tends to increase. In extreme cases, the circuit could go into thermal runaway, with the rising current potentially destroying the output transistors.
The solution is to incorporate into the bias circuit a component that produces the opposite effect, maintaining a stable idle current over the amplifier’s entire operating temperature range — a task easier said than done. In practice, a diode is placed in physical contact with the heat sink to capture its temperature and connected to the bias adjustment circuit to provide a reference, ensuring idle current stability.
Interestingly, MOSFETs have a positive temperature coefficient and thus do not require a compensation circuit, though they introduce other challenges.
Up to this point, everything seems fine, but certain characteristics of Class AB amplifiers can still prevent them from achieving the performance level of Class A amplifiers. Among other issues, the inherent linearity of a push-pull stage is unfortunately inferior to that of Class A, especially for low-intensity signals, as the transistors in this case do not operate within their ideal range. Nonetheless, a carefully designed circuit with a reasonable amount of feedback can deliver very impressive results. Another key factor affecting performance is the power supply.
As mentioned earlier, one advantage of Class A is its constant current demand, resulting in a stable power supply voltage that supports dynamic sound reproduction. In Class AB, the current demand varies with the intensity of the audio signal, from a few tens of milliamps to several amps, causing fluctuations in the power supply voltage that may lead to less lively performance. The solution is to oversize the power supply section or regulate its output to ensure stability, though this significantly impacts manufacturing costs.
Nevertheless, the performance level achieved by the best-designed Class AB devices is comparable to that of Class A, at a lower cost and without the thermal drawbacks.
We previously noted that some Class AB amplifiers have output stage bias levels much higher than the few milliamps mentioned earlier. While Class AB offers significantly better pre-feedback linearity than Class B, the region immediately following the start of conduction in the output transistors does not exhibit ideal performance.
For this reason, many manufacturers use a higher idle current, around 100 mA, to ensure that the transistors never leave their optimal operating range. In some cases, the bias is set even higher, allowing the amplifier to operate in Class A for the initial watts. For instance, with an idle current of 1 A, the first 8 watts into an 8 Ω load would effectively operate in Class A. Naturally, the idle thermal dissipation would be higher, and using our earlier example of an amplifier with ±40 VDC power rails, it would reach 80 W at idle for a maximum output power of just under 100 W RMS. This efficiency loss is significant and requires substantial heat sinks and a more refined bias control circuit.
All the circuits discussed thus far have used the output stage exclusively as a current amplifier, even introducing a slight voltage loss. In addition, it is worth mentioning amplifiers whose output stages provide voltage gain in addition to current gain. This concept is interesting because it not only offers a power gain (P = V x I) but also allows for simpler voltage amplification stages, which is advantageous for performance without increasing manufacturing costs.
The following diagram shows such a circuit. At first glance, apart from the addition of two transistors before the final output stage, it appears nearly identical to the previous one. However, note that the NPN and PNP transistors are reversed. As a result, the speaker is now connected to the collectors of the transistors rather than the emitters. Referring to the various configurations discussed in the first article of this series, we see that instead of operating in common-collector mode, where only current gain is possible, the circuit now operates in common-emitter mode, offering both voltage and current gain.
Class AB Amplifier with Voltage Gain at the Output Stage
Next month, we will review less commonly used operating classes, paying particular attention to Class D. As we will see, this technology has reached a level where it is becoming a serious competitor to traditional configurations in terms of reproduction quality, while offering significant advantages that should soon make it the go-to choice for HiFi systems. Stay tuned…
Let’s talk technical to be continued soon…
